Keynotes & Special Sessions

Physical AI systems must perceive, predict, and safely interact with dynamic real-world environments. This keynote presents our vision of a real-time digital twin platform as a real-world foundation integrating sensing, communications, computing, AI, and control. We first introduce lessons learned from the NSF-NICT-JUNO project, in which mobility digital twins were developed and experimentally validated for connected and automated driving. Building on these achievements, the JST ASPIRE project, involving ten institutions from seven countries, aims to establish an internationally standardized common platform for developing, training, validating, and operating Physical AI across mobility, industry, agriculture, and nursing care. Demonstrations in mobility and agriculture will illustrate the platform’s cross-domain applicability and its ability to share real-world data, digital twin models, AI capabilities, and computational resources across different fields. Finally, we discuss the future integration of humanoids as common physical actuators capable of acquiring and performing diverse tasks across all these application fields, paving the way toward an open, interoperable, and globally shared real-world foundation for Physical AI.

Kei Sakaguchi (Senior Member, IEEE) received the M.E. degree in information processing and the Ph.D. degree in electrical and electronics engineering from Tokyo Institute of Technology, Japan, in 1998 and 2006, respectively. Currently, he is the Program Director (VI-ILS), the Dean of the Academy of Super Smart Society, and a Professor with the School of Engineering, Institute of Science Tokyo (formerly Tokyo Institute of Technology). His current research interests include 5G/6G cellular networks, millimeter-wave communications, V2X for automated driving, digital twins, smart mobility, smart agriculture, smart ocean, and smart healthcare. He is a fellow of IEICE. He received the Outstanding Paper Awards from SDR Forum and IEICE, in 2004 and 2005, respectively, and the three Best Paper Awards from the IEICE Communication Society, in 2012, 2013, and 2015. He also received the Tutorial Paper Award from the IEICE Communication Society, in 2006.

Due to the explosive growth in the number of wireless devices and diverse wireless services, next-generation wireless networks face unprecedented challenges caused by heterogeneous data traffic, massive connectivity, ultra-high bandwidth efficiency and ultra-low latency requirements. To address these challenges, flexible-antenna systems have been recognized as key enabling technologies of the sixth-generation (6G) wireless networks, as they can intelligently reconfigure users’ effective channel gains and hence significantly enhance their data transmission capabilities. However, the existing flexible-antenna systems have been developed to combat small-scale fading in non-line-of-sight (NLoS) conditions. As a result, they are lack of the capability to reconstruct strong line-of-sight (LoS) links which are typically 100 times stronger than NLoS links. Furthermore, the existing flexible-antenna systems exhibit restricted flexibility, where adding/removing an antenna is not straightforward. This talk focuses on an innovative flexible-antenna system, termed generalised pinching-antenna systems, and how such a new physical technology can reshape the future network architecture. The principles of generalized pinching-antenna systems are described first together with specific examples of generalized pinching-antenna systems, including Docomo’s dielectric waveguide based pinching antennas, leaky coaxial cable (LCX), etc. In addition, promising 6G related applications of generalized pinching antennas, including environment division multiple access (EDMA), integrated sensing and communication, multi-cell interference management, etc, are also illustrated. Finally, important directions for future research, such as antenna/waveguide deployment, channel estimation, etc, are highlighted.

Zhiguo Ding is currently a Professor in Communications at Nanyang Technological University and an Academic Visitor at Princeton University. His research interests are 6G networks, communications and signal processing. His h-index is over 130 and his work receives 80,000+ Google citations. He is serving as the EiC of IEEE JSAC, an Area Editor for the IEEE TWC and OJSP, an Editor for IEEE TVT, and OJ-SP, and was an Area Editor for IEEE TCOM and OJCOMS, an Editor for IEEE TCOM, TWC, COMST, WCL, CL and WCMC. He received the best paper award of IET ICWMC-2009 and IEEE WCSP-2014, the EU Marie Curie Fellowship 2012-2014, the Top IEEE TVT Editor 2017, IEEE Heinrich Hertz Award 2018, IEEE Jack Neubauer Memorial Award 2018, IEEE Best Signal Processing Letter Award 2018, Alexander von Humboldt Foundation Friedrich Wilhelm Bessel Research Award 2020, IEEE SPCC Technical Recognition Award 2021, IEEE VTS Best Magazine Paper Award 2023, and the Best Paper Award in IEEE GLOBCOM 2024. He is a Web of Science Highly Cited Researcher in two disciplines (2019-2025), and a Fellow of the IEEE.

Open Radio Access Network (O-RAN) has emerged as a promising paradigm for realizing intelligent, open, and programmable radio access networks. Despite its potential, significant challenges remain in improving network energy efficiency and optimizing end-to-end service performance. The Non-Real-Time RAN Intelligent Controller (Non-RT RIC) provides a powerful framework for addressing these challenges through AI/ML-driven automation and network optimization. At the same time, the integration of terrestrial networks (TNs) and non-terrestrial networks (NTNs) has attracted growing interest as a means of extending network coverage and ensuring seamless quality-of-service (QoS) provisioning.

To address energy-efficiency requirements, we previously proposed an energy-efficient O-RAN power management architecture that dynamically controls base station (BS) activation and deactivation using AI/ML while leveraging NTN resources to maintain service continuity. Our investigation revealed that, while such an approach is effective for local optimization, broader end-to-end optimization across heterogeneous networks requires coordination among multiple control domains. This limitation extends beyond energy-efficient network operation and also affects traffic steering, resource management, and distributed AI lifecycle management across heterogeneous network domains.

In this presentation, we introduce a vision, architecture, and operational framework for a multi-Non-RT RIC orchestration system that enables network-wide energy-efficient management, cross-domain resource optimization, intelligent traffic steering, and distributed AI lifecycle management. Building upon this vision, we provide an overview of our research and development activities. First, we review research efforts on energy-efficient power management for O-RAN base stations by jointly utilizing AI/ML and NTN resources, and discuss future research directions. Second, we present recent advances in TN-NTN convergence technologies, including architectures and operational scenarios for integrated and distributed network control systems that support end-to-end service design and orchestration, performance monitoring, and resource management across both TN and NTN environments. In addition, we introduce joint demonstration activities conducted by the National Institute of Information and Communications Technology (NICT) and the Singapore University of Technology and Design (SUTD), which were exhibited at Mobile World Congress (MWC) 2025 and MWC 2026. These demonstrations highlight the practical feasibility and benefits of intelligent, energy-efficient, and converged network management based on O-RAN, NTN, and AI/ML technologies.

Takaya Miyazawa received his M.E. and Ph.D. degrees in Information and Computer Science from Keio University, Yokohama, Japan, in 2004 and 2006, respectively. From April 2006 to March 2007, he was a Visiting Researcher at the University of California, Davis, CA, USA. He joined the National Institute of Information and Communications Technology, Tokyo, Japan, as a Researcher in 2007 and is currently a Research Manager. From April 2019 to July 2020, he served as a Deputy Director at Japan’s Ministry of Internal Affairs and Communications before returning to NICT in August 2020.

He has contributed to the IEEE Communications Society Asia-Pacific Region as an officer for more than 12 years and currently serves as Co-Chair of the Information Services Committee. His research interests include network control and management. He received the Hiroshi Ando Memorial Young Engineer Award in 2007, the Funai Young Researcher Award in 2010, the Best Paper Award at the ITU Kaleidoscope Academic Conference in 2018, and the IEICE Distinguished Contributions Awards in 2020, 2021, and 2024. He is a member of the IEICE, the IEEE and the IEEE Communications Society.

Earth observation is generating rapidly increasing volumes of multimodal data from optical, SAR, hyperspectral, LiDAR, and other sensing platforms. At the same time, increasingly capable foundation models are creating new opportunities for large-scale geospatial understanding. However, operational deployment remains challenging because Earth-observation AI must adapt across sensors, geographic regions, resolutions, and downstream tasks while operating under constraints in labelled data, computation, communication, and onboard resources. This talk will introduce our recent work towards Efficient Space AI that make Earth-observation intelligence more scalable, adaptive, and deployable.

The talk will cover efficient multimodal learning, reusable foundation representations, data- and parameter-efficient adaptation, spatial-temporal intelligence, and lightweight edge and onboard inference. We will also discuss how these capabilities can support systems that integrate heterogeneous observations, reason over changes across space and time, and provide traceable evidence for downstream applications. Examples from satellite-based environmental intelligence will illustrate how Efficient Space AI can bridge advances in foundation models with practical and scalable Earth-observation applications.

Dr. ZHU Hongyuan is the Principal Scientist and PI of Satellite Sensing Team at the Agency for Science, Technology, and Research (A*STAR), Singapore. He is leading the Advanced Perception and Reasoning Lab, focusing on developing autonomous agents with super-large-scale multimodal sensing and reasoning capabilities for sustainability, climate/weather, and defence solutions. He was selected as the top 2% scientist by Stanford in 2023~2025. He was the recipient of A*STAR Career Award in 2022. His team won the 1st place winner of Scene2Cap challenge ICCV23, 3rd place of EPIC Text-Video Retrieval Challenge in CVPR22 and the only Asia team as 1st Prize Finalist of KUKA Innovation Challenge 2021. He is currently the associate editor of Visual Computer since 2020. He also served as the Senior Program Committee member of IJCAI, the Area Chair of ACM MM Asia, and the Guest Editor of IET Image Processing. He has published around 120 top-tier journal and conference papers, such as CVPR, ICCV, NeurIPS, ICML, AAAI, IJCAI, ACL, TPAMI, etc.

Device-free localization techniques using channel state information (CSI) in wireless local area networks (WLANs) have been extensively studied. In these techniques, the collected CSI is used as a dataset for machine-learning (ML)-based localization. This talk presents experimental results and comparisons of a recently developed WLAN-based device-free localization scheme under different environmental scenarios. The results demonstrate that the effectiveness of the localization scheme depends on the placement of WLAN transceivers and the presence of reflective obstacles near the experimental area.

Osamu Muta received the Associate Degree in Engineering from Sasebo Institute of Technology in 1994, the B.E. degree from Ehime University in 1996, the M.E. degree from Kyushu Institute of Technology in 1998, and the Ph.D. degree from Kyushu University in 2001. In 2001, he joined the Graduate School of Information Science and Electrical Engineering, Kyushu University, as an Assistant Professor. From 2010 to 2023, he was an Associate Professor at the Center for Japan-Egypt Cooperation in Science and Technology, Kyushu University. Since 2023, he has been a Professor at the Faculty of Information Science and Electrical Engineering, Kyushu University, Japan. His research interests include signal processing for wireless and power line communications, MIMO techniques, interference coordination, wireless sensing, low-power wide-area networks, and nonlinear distortion compensation for high-power amplifiers.

In this talk, I introduce the potential of Deep Joint Source-Channel Coding (DeepJSCC) for underwater networks. DeepJSCC directly transmits latent image representations as IQ signals and is well suited to challenging underwater channels. We demonstrated its effectiveness in underwater acoustic, underwater optical, and 5G systems, and achieved real-time FPGA operation at approximately 33 fps with low power consumption.

Daisuke Hisano received the B.E., M.E., and Ph.D. degrees from Osaka University, Osaka, Japan, in 2012, 2014, and 2018, respectively. Since April 2026, he has been an Associate Professor at Keio University. His research interests include AI-driven communications, optical communications, and all-optical signal processing.

This talk presents a novel approach to address the severe performance degradation caused by rapidly varying Doppler shifts and fractional Doppler effects in orthogonal time-frequency space (OTFS)-based vehicle-to-everything (V2X) communications. Focusing on high-speed counter-approaching scenarios, we highlight the vulnerability of conventional OTFS receivers when continuous Doppler variation disrupts channel sparsity. To overcome this challenge, we introduce a dual-branch receiver architecture that processes incoming signals in parallel—one branch with no offset and another with an intentional half-grid Doppler shift. By independently performing channel estimation and equalization, and dynamically selecting the branch with the highest post-equalization signal-to-interference ratio (SIR), our method provides a robust and low-overhead solution. Simulation results validate that this approach significantly mitigates fractional Doppler-induced degradation, offering substantial throughput improvements over conventional designs in realistic V2V environments.

Dr. Keiichi Mizutani is a Professor in the Department of Information and Communication Engineering, Graduate School of Engineering, Nagoya University, Japan. He received the B.E. degree from Osaka Prefecture University in 2007, and the M.E. and Ph.D. degrees from Tokyo Institute of Technology in 2009 and 2012, respectively. From 2009 to 2012, he was a Research Fellow of the Japan Society for the Promotion of Science (JSPS), and in 2010 he was a visiting researcher at the Fraunhofer Heinrich Hertz Institute, Germany. He joined the National Institute of Information and Communications Technology (NICT) as a researcher in 2012. In 2014, he moved to the Graduate School of Informatics, Kyoto University, where he served as an Assistant Professor, a Program-Specific Associate Professor, and an Associate Professor before joining Nagoya University in 2025.

His research interests include wireless communication systems and modulation, cognitive radio, software-defined radio, dynamic spectrum access, TV white space communications, full-duplex communications, and machine learning for wireless communications. His current work focuses on OTFS for high-mobility vehicular communications, communication control for dense three-dimensional drone networks, and radio environment monitoring for industrial sites. He contributed to the standardization of IEEE 802.15.4m, IEEE 802.11af, and IEEE 802.22b, receiving an IEEE Standards Association certificate of appreciation for IEEE 802.15.4m in 2014.

His awards include Best Paper Awards at WPMC 2014, 2017, 2020, and 2022 and IEEE LANMAN 2017, and the IEEE WF-IoT 2020 Best Paper Presentation Award. He served as Chair of the IEICE Technical Committee on Short Range Wireless Communications from 2023 to 2025 and is currently Vice Editor-in-Chief of IEICE Communications Express. He is a member of IEEE and IEICE.

Stacked intelligent metasurfaces (SIMs) have recently emerged as a promising technology for manipulating electromagnetic waves directly in the wave domain. By stacking multiple programmable metasurface layers, SIMs can provide richer spatial transformations than conventional single-layer intelligent surfaces, opening new possibilities for the design of future wireless communication systems. This talk provides an overview of SIMs from the perspectives of both electromagnetic wave propagation and wireless communications. We first introduce the basic concept and operating principles of SIMs, with particular attention to how electromagnetic waves propagate through and between multiple metasurface layers. We then discuss several approaches to modeling such propagation and highlight the importance of appropriately capturing the physical characteristics of stacked structures. While simplified propagation models are useful for communication-theoretic analysis and optimization, practical SIM implementations may involve various electromagnetic effects, such as propagation loss, coupling, and multiple interactions among metasurface layers. These effects can influence the effective signal transformation realized by SIMs and consequently their achievable communication performance. The talk will also introduce recent research directions and selected examples of SIM-assisted wireless communication systems, including wave-domain signal processing and advanced multiple-access techniques. Through these examples, we discuss how the physical characteristics of SIMs can be incorporated into communication-system design and optimization. Finally, we briefly present emerging applications and open research challenges toward practical SIM-enabled wireless systems.

Hiroaki Hashida is an assistant professor with the Frontier Research Institute for Interdisciplinary Sciences, Tohoku University, Sendai, Japan. He received his Ph.D. degree in information science from Tohoku University in 2024. He was a recipient of the Presidential Award for Outstanding Students from Tohoku University and the Ikushi Prize from the Japan Society for the Promotion of Science (JSPS) in 2024. His research interests include wireless communication networks, intelligent surface-aided wireless communication systems, and stacked intelligent metasurfaces. His research focuses on the design and optimization of future wireless communication systems by integrating advanced wireless technologies with programmable and reconfigurable radio environments. In particular, he is interested in developing efficient methods for controlling and utilizing intelligent surfaces to improve the coverage, reliability, and efficiency of wireless networks. He is a member of the IEEE and the Institute of Electronics, Information and Communication Engineers.

In this talk, we will introduce two innovative types of 3D intelligent metasurfaces: Stacked Intelligent Metasurfaces (SIM) and Flexible Intelligent Metasurfaces (FIM). We will explore their exciting applications in wireless communication and sensing systems. Specifically, SIM is a groundbreaking computing architecture that enables joint signal processing and communication in the electromagnetic (EM) domain. A SIM is fabricated by stacking an array of programmable metasurface layers, where each layer consists of many low-cost passive meta-atoms that can individually manipulate EM waves. By appropriately configuring the passive meta-atoms, a SIM can automatically accomplish advanced computation tasks as the EM wave propagates through it while reducing both energy consumption and processing delay. By contrast, an FIM takes a different approach to leverage the 3D physical space. An FIM is composed of an array of low-cost radiating elements, each of which can independently radiate electromagnetic signals while flexibly adjusting its position along the direction perpendicular to the surface. Hence, unlike conventional rigid 2D antenna arrays, the FIM surface shape may be dynamically reconfigured to improve the channel quality by beneficial 3D morphing.

Chau Yuen received the B.Eng. and Ph.D. degrees from Nanyang Technological University, Singapore, in 2000 and 2004, respectively. Since 2023, he has been with the School of Electrical and Electronic Engineering, Nanyang Technological University. Dr. Yuen received IEEE Communications Society Leonard G. Abraham Prize (2024), IEEE Communications Society Best Tutorial Paper Award (2024), IEEE Communications Society Fred W. Ellersick Prize (2023), IEEE Marconi Prize Paper Award in Wireless Communications (2021), IEEE APB Outstanding Paper Award (2023), and EURASIP Best Paper Award for JOURNAL ON WIRELESS COMMUNICATIONS AND NETWORKING (2021). He is an IEEE Fellow and also a Highly Cited Researcher by Clarivate Web of Science.

Detection of illegal and unwanted UAVs is in high demand for air safety and security in civil society. Conventional optical imaging offers cost-effective and attractive solutions; however, adverse weather conditions such as fog and darkness prevent successful detection. Millimeter-wave radar systems provide another possible solution, but certain UAV shapes and orientations can dramatically degrade the reflected signal strength. We are currently developing a multistatic imaging system combining synchronized millimeter-wave radars with multilateration-based detection, leveraging both communication and radar sensing techniques. In this talk, we will briefly introduce the overview and principles of the system, with a particular focus on wireless clock and frequency synchronization techniques.

Mr. Kunihiko Sasaki is a visiting professor at the Nagoya Institute of Technology. For over 40 years, he has conducted research on technologies such as wireless communication, antenna propagation, and wireless power transfer at DENSO Corporation's research laboratories. Concurrently, he has spearheaded standardization efforts related to the Japanese government's Radio Act.

The growing demands of satellite sensing and non-terrestrial networks call for efficient onboard processing, low-latency inference, and reduced dependence on resource-intensive digital computation. Conventional satellite remote-sensing systems typically follow a “digitize-then-process” paradigm, where high-dimensional sensing data are digitally processed onboard or transmitted to terrestrial stations for subsequent reconstruction and semantic inference, resulting in substantial computational and communication overhead.

This talk explores an alternative “compute-while-transmitting” paradigm that exploits electromagnetic wave propagation as a physical computational process. As a representative case study, we present a stacked intelligent metasurface-based diffractive neural network for onboard terrain classification directly from Synthetic Aperture Radar (SAR) Level-0 raw data. Multiple programmable metasurface layers are jointly optimized to perform task-oriented feature mapping in the wave domain, while received signal intensities at the terrestrial station directly provide semantic classification outputs. A lightweight phase-domain augmentation strategy is further introduced to improve the learnability of complex-valued raw SAR signals under speckle, Doppler distortions, and other impairments.

Numerical results demonstrate approximately 90% classification accuracy for binary terrain recognition directly from real SAR Level-0 data, while reducing reliance on conventional digital processing and high-dimensional raw-data transmission. More complex multi-class experiments further reveal the representational limitations of purely linear in-wave processing, motivating nonlinear and hybrid physical computing mechanisms.

Beyond SAR classification, this work highlights the potential of programmable electromagnetic structures to evolve from communication components into task-oriented physical computing platforms, opening opportunities for integrated sensing, communication, and computation in future terrestrial and non-terrestrial network architectures.

Mengbing Liu received the B.E. degree in Electronic Information Engineering from Northeastern University, China, in 2018, and the M.E. degree in Electronics and Communications Engineering from the University of Science and Technology of China, China, in 2021. She is currently pursuing the Ph.D. degree with the School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore.

Her research focuses on physics-driven machine learning and intelligent wireless systems, with particular interests in stacked intelligent metasurfaces, reconfigurable intelligent surfaces, in-wave and analog computing, and integrated communication, sensing, and computation. Her recent work explores programmable electromagnetic structures for physical-layer inference and task-oriented processing in next-generation wireless and satellite systems. Her research has appeared in leading venues including IEEE TWC, IEEE TCCN, and IEEE Wireless Communications, among others.

In recent years, various autonomous mobile robots (AMRs) such as automated guided vehicles (AGVs), food-serving robots in restaurants, and remotely operated guidance robots—have begun to spread daily life. These AMRs transmit sensor data, which are measured by sensors mounted on the units, to a control server, and they are controlled remotely by commands received from the server. Since these tasks must be performed while the autonomous mobile units are moving, information exchange between the units and the server is performed through wireless communications.

Autonomous mobile units are equipped with CPU boards for control purposes. Some units also have components such as displays to provide interactive services to people. CPU boards and communication lines emit radio-frequency noise—spanning frequencies up to several gigahertz. It can cause significant interference with nearby wireless communications.

The Ministry of Internal Affairs and Communications is conducting commissioned research “Research and development on technologies for analysis and suppression of unwanted radio waves between densely-deployed mobile communication devices” in research and development for expansion of radio wave resources (JPJ000254). In this commissioned research, unwanted radio waves emitted from electronic devices such as small computers is measured, and its impact on the performance of several communication systems including the IEEE 802.11ah (IEEE 802.11 S1G) system and LTE/5G systems is evaluated. Furthermore, the methods to suppress the unwanted radio waves by using transparent flexible radio-wave reflective film are studied.

In this commissioned research, a real-time radio monitoring system which can monitor spectrum utilization in time and frequency domains simultaneously. This monitor can capture instantaneous spectrum utilization and behaviors of wireless communication systems. For example, the collisions in channel access between IEEE 802.11 S1G and IEEE 802.15.4g (IEEE 802.15.4 SUN) systems were observed by this monitor.

This talk introduces the activities of this commissioned research.

Dr. Kazuto Yano received the B.E. degree in electrical and electronic engineering, and M.S. and Ph.D. degrees in communications and computer engineering from Kyoto University in 2000, 2002 and 2005, respectively. He was a research fellow of the Japan Society for the Promotion Science (JSPS) from 2004 to 2006. Since April 2006, he has been a researcher at Advanced Telecommunications Research Institute International (ATR). Currently, he is the head of Department of Wireless Communication Systems at Wave Engineering Laboratories, ATR.

His research interests include space-time signal processing for interference suppression, MIMO transmission, and PHY/MAC cross-layer design of wireless communication systems for license-exempt bands. He received IEICE (the Institute of Electronics, Information and Communication Engineers) Communications Society Best Tutorial Paper Award in 2017, and ICAIIC 2019 Excellent Paper Award in 2019. He is a senior member of IEICE, and a member of IEEE. He serves as IEEE 802.19 Working Group Editor and a Vice Chair of IEEE 802.19 Task Group 3a (TG3a).

With the rapid advancement of computing power, Artificial Intelligence (AI) has become an increasingly valuable assistant in our daily lives. However, the adoption of AI in engineering applications remains relatively limited. In this technical sharing session, we will introduce an integrated 3-in-1 Engineering AI solution that combines Physics AI, Agentic AI, and Generative AI. Through practical applications and real-world case studies, we will explore how these three AI technologies can be leveraged to address engineering challenges, enhance productivity, and drive innovation in engineering workflows.

Dr. Vincent Chai is a Senior Consultant at CADFEM with nearly 20 years of experience in Computational Fluid Dynamics (CFD). In his previous roles, he served as Deputy Principal Engineer at NEA (EHI), where he conducted CFD analyses and field tests related to indoor air quality applications. He was also an Associate Director at BECA, leading the CFD programme for building HVAC analyses.

Additionally, he worked as a Senior Consultant at an ANSYS channel partner company, supporting clients from various industries on advanced CFD applications.

Earlier in his career, he was a Research Engineer at NIDEC, conducting research on the aerodynamic performance of wind turbines and fan products, and a Project Associate at Nanyang Technological University, working on blast-related applications.